Is it needed or correct to use -regress_reml_exec option in surface-based analysis?

AFNI Message Board
1

Hi,

Is it needed or correct to use -regress_reml_exec option in surface-based analysis( like afni_proc example 8)?
If this option is added, there is an error when calculating TSNR after regression. The proc script use errts.${subj}.$hemi.niml.dset"[$ktrs]" in this step, while its real name is errts.${subj}.$hemi.niml.dset_REML.niml.dset"[$ktrs]“, so there will be a " FATAL ERROR”.

Thank you,
2086

2

Hi 2086,

Yes, that should be as appropriate and useful as in the volume.

I will try to check out the problem with afni_proc.py fairly soon.
Okay, it looks like we need to make little updates to both
afni_proc.py and 3dDeconvolve (the program that write the
3dREMLfit script). Until that is fixed, you can modify the
‘noise’ dataset in the proc scripts by appending, _REML.niml.dset

  • rick
3

Hi Rick,

Thank you for your reply.

Appending _REML.niml.dset to proc scripts does work, but in the final review part, gen_ss_review_scripts.py is still not fully compatible with REML, it can’t find “final stats dset”, and @ss_review_basic doesn’t report “TSNR average”.

2086

4

Hi Rick,

This problem still exists in the latest version of AFNI(Dec 11 2018).
The proc script call for errts.${subj}.lh.niml.dset"[$ktrs]" while it is named as errts.${subj}.lh_REML.niml.dset"[$ktrs]".
If you run afni_proc with -regress_3dD_stop, you can find this problem.

Thank you,
2086